Registration Request: meta

[marrus-sh]

Relation Name

meta

Description

Links to a resource containing metadata about the link’s context

Reference

https://www.w3.org/TR/cooluris/#linking

Additional Information

This has been existing practice on the Web for some time. The provided reference is the most formal one to my knowledge (§4.6 Linking, near the bottom, describes rel="meta" as an alternative to rel="alternative" when the metadata resource merely describes, and does not attempt to provide an alternative for, the link’s context). But rel="meta" is also described in:

• The W3C Wiki page for the HTTP Link Header
• The Microformats wiki
• The FOAF specification, in the section on autodiscovery

[dret]

Hello.

On 2024-08-28 09:29, kibigo! wrote: The provided reference is the most formal one to my knowledge (§4.6 Linking, near the bottom, describes |rel="meta"| as an alternative to |rel="alternative"| when the metadata resource merely describes, and does not attempt to provide an alternative for, the link’s context).

That latter definition ("the metadata resource merely describes [...] the link's context") sounds exactly what the registered "describedby" link relation type is defining. Is there any difference? If yes, it might be useful to spell that out. If not, it may be useful to clearly say that "meta" is a synonym for "describedby" and that they can be treated interchangeably. Cheers, dret.

…

-- Erik Wilde | ***@***.*** | | https://youtube.com/ErikWilde |

[mnot]

This is a third-party registration of a value that's mentioned briefly in a very old W3C Interest Group note - any thoughts @plehegar?

I also note that the reference doesn't define the semantics; instead, the registration infers them.

[marrus-sh]

That latter definition ("the metadata resource merely describes [...] the link's context") sounds exactly what the registered "describedby" link relation type is defining. Is there any difference?

The practical answer here is that describedby was defined by POWDER in 2009, and my expectation is that RDF autodiscovery tools (the primary use‐case) which are not POWDER‐specific are much, much more likely to look for rel="meta" than rel="describedby". It’s older, it’s what the documents cited above use, etc…. So they may very well be equivalent terms, but in practice I worry that existing tools probably treat them differently (i.e. recognize one but not the other).

Anyway, in case it wasn’t clear, I think the “specification” of rel="meta" leaves a lot to be desired. My understanding (I was very young in 2008 when these conversations were actually happening) is that it mostly came about through discussions on mailing lists and wikis and the like, and nobody ever bothered to formally specify it. It is also hampered by the fact that the primary use case seems to be with HTML <link> elements, and HTML has a lot looser understanding of link relations which doesn’t require registration or clear specification. However, as the W3C Wiki page shows, people have also used it with Link headers for resources which are not HTML.

The HTML5 link type extensions page cites an old version of RDF/XML as its source for rel="meta", which in turn cites RFC2731 (and attributes the choice to DCMI). But RFC2731 doesn’t actually define rel="meta" anywhere as far as I can tell. This section was removed in the latest version of the RDF/XML specification, for reasons which are not explained (but may have something to do with the transition from old XHTML to (X)HTML5).

In any case, I think the above history means it would be very problematic if "meta" ever were to be assigned to some purpose other than that defined above. That, to me, is a pretty strong argument in favour of registration; the reality is that this value is and has been used for over a decade.